Public Abstract

DE-SC0012516: Optimizing the Energy Usage and Cognitive Value of Extreme Scale Data Analysis Approaches

Award Status: Inactive

Institution: The University of Texas at Austin, Austin, TX
UEI: V6AFQPN18437
DUNS: 170230239

Most Recent Award Date: 11/22/2017
Number of Support Periods: 3
PM: Biven, Laura

Current Budget Period: 09/15/2016 - 09/14/2018
Current Project Period: 09/15/2014 - 09/14/2018
PI: Samsel, Francesca

Supplement Budget Period: N/A

Public Abstract

Optimizing the Energy Usage and Cognitive Value of Extreme Scale Data Analysis Approaches James Ahrens (large scale data analysis) LANL, Wu Feng (energy) Va Tech, Colin Ware (cognition/perception), Greg Abram (algorithms) UT Austin, Francesca Samsel (perception) UT Austin The goal of this project is to understand how to maximize the scientific insight gained from extreme scale simulation results, while at the same time minimizing power consumption needed to compute, analyze and store data. To achieve this, the research team brings together experts in perception and human cognition with experts in extreme data analysis, supercomputing energy usage, and scalable algorithms. The team will follow an experimental approach to adaptively sample extreme scale data as it is being computed, and then - through experiments in perceptual and cognitive science - evaluate the impact these sampling algorithms have on the results that are stored. The team expects to deliver new open source data analysis algorithms that can be adapted to a wide range of workflows. Resulting measurements and models will ideally help to develop a community-wide framework and approach to significantly increasing the cognitive value of data from extreme scale simulations.

Optimizing the Energy Usage and Cognitive Value of

Extreme Scale Data Analysis Approaches

James Ahrens (large scale data analysis) LANL, Wu Feng (energy) Va Tech, Colin Ware (cognition/perception), Greg Abram (algorithms) UT Austin,

Francesca Samsel (perception) UT Austin

The goal of this project is to understand how to maximize the scientific insight gained from extreme scale simulation results, while at the same time minimizing power consumption needed to compute, analyze and store data. To achieve this, the research team brings together experts in perception and human cognition with experts in extreme data analysis, supercomputing energy usage, and scalable algorithms. The team will follow an experimental approach to adaptively sample extreme scale data as it is being computed, and then - through experiments in perceptual and cognitive science - evaluate the impact these sampling algorithms have on the results that are stored. The team expects to deliver new open source data analysis algorithms that can be adapted to a wide range of workflows. Resulting measurements and models will ideally help to develop a community-wide framework and approach to significantly increasing the cognitive value of data from extreme scale simulations.